Hacker News Reader

Comment by freedomben | Hacker News Reader

Yeah this has been a huge problem for me too. I'm about to just go Intel instead. I hope they ramp up production and don't miss out because I think a strong AMD would be a good thing. I can't wait forever to build my new rig though and I've been using Intel for decades so I have confidence in it. Excited to go AMD but I am building a new rig for a reason (my old one is in its death throes and I use this thing all day almost every day for work).

View on HN

>I hope they ramp up production

Every high-performance CPU released lately uses TSMC for manufacturing, which can't keep up with demand.

See Playstation 5/Xbox shortages, Ryzen and Radeon shortages. I think only Apple was able to secure enough manufacturing.

Would Nvidia's shortages also fall in this bucket?

Nvidia used samsung. Rumor has it that their shortage is because they sold their cards directly to cryptominers.

I believe this was debunked. The sales to cryptominers only represents a few percents of the total.

That is stuff angry gamer kids make up to find a reason to be mad. The reality is that manufacturing is generally slowed due to covid, and demand is generally high due to covid, and nobody can keep up.

Is mining on GPUs still something people do? Or can be done anywhere near profitable? I got the impression with ASICs that game was up.

It's for ASIC resistant currencies I believe

Nvidia's RTX 3000 line is manufactured by Samsung, not TSMC.

We know more about Nvidia shortage on 3000 series: Samsung wafer, substrates and components (maybe the GDDR6X memory chip which only Micron can make).

[1] https://www.techpowerup.com/275530/nvidia-rtx-30-series-shor...

All rumors about low supply 3000 series are debunked by looking at their quarterly revenue results and the Steam ranking that was released yesterday. (See my other reply.)

If you don't need the best single core performance, the cheaper Ryzen 3000 chips are still available.

And unlike intel they've been using the same sockets for the last 3 generations, so upgrading the CPU in future is actually a reasonable idea.

My old desktop is still a beast with a 10 core Xeon and 128GB of ram. I'm investing in a new rig to speed up my rust compiles. The current machine is 4 years old now, so I can afford to wait a few months for a build that will last me that long.

Do you actually get real RoI on that thing? Must have cost magnitudes more than normal desktop hardware. Or is this where you just accept that it's worth it purely for tingling the geek inside you?

A faster compiling machine is the difference between sustaining a state of flow critical to developer productivity, or not. Paying even $5000 for a desktop that yields probably a 2x improvement in productivity for someone paid at least $10K a month is a total no-brainer.

I think 2x improvement in productivity is a pretty optimistic number unless you don't write much code and mostly just compile it. In heavier workloads there certainly can be decent ROI though.

Compile time impact is a very non-linear thing. There's a narrow band that could be called the distraction threshold. An investment in faster hardware will give very small returns if you are already below that range and only moderate returns when so far above that the investments only gets you a little closer. But when the speedup crosses that narrow band or at least significant parts of it, it's a night and day difference.

Is there an average number for the distraction threshold or is each person pretty different?

Once it goes over about twenty seconds for me, I'm switching tasks or checking my email (or worse, HN.)

Keeping that compile time down is critical to being productive.

And the 24 Core 128G RAM machines are how we get bloated software...

(not blaming anyone in particular)

My dad got me a really good deal ($600!) on the RAM on ebay, otherwise I think 64gb would have been plenty. The nice thing about it is the Linux buffer cache keeps everything in memory, the SSD is just for durability.

Over 4 years of heavy software development I think the system has paid for itself in time saved.

I built a TR 3960X (24 core) 32GB of RAM workstation for $3000, including a 700 dll graphics card (2080 Super). Realistically a 100 dollar graphics card would've sufficed for my usage (development). So it could've been built for $2400.

Depending on the context such a machine could be very cheap or very expensive, but even here in Mexico that's about one month of an engineer's salary, which I don't find too outrageous for a primary tool which lasts for years.

Depending on your workload, the performance improvements could be very noticeable (running multiple VMs, compiling, and unit testing at the same time). It does open up faster workflows if you adjust to the new speed.

Old server hardware can be very cheap, although it's not necessarily the best value. Typically it only makes sense if 1) you need a lot of memory, 2) you don't mind a slightly higher power bill and 3) you don't care about single core performance.

I've used a 12-core Haswell Xeon as a desktop for about five years (got it used for around $500, plus $200 for a brand new LGA 2011 v3 motherboard). At the time it was twice as fast as a high-end desktop 4-core CPU. It's still reasonably competitive on multithreaded workloads versus a recent 6 core desktop CPU, but fares poorly on single-threaded tasks. DDR4 was a bit expensive back then, but quad channel memory helps in that you can buy lower density DIMMs. Today, even DDR4 ECC RDIMM server pulls are dirt cheap. I picked up 4x16GB DDR4-2400 RDIMMs for just $90 last year.

I'll probably replace it with a 5900X soon, but it's served well over the past five years. Certainly much better than a i7-4770K box from the same time.

If you can hang on for a couple more weeks. Supply side is supposed to get a lot better on the AMD cpus by the last two weeks of december.

Reply on news.ycombinator.com